Online joint localization without user interactions

Abstract

Wi-Fi fingerprinting has been a popular indoor localization due to the widespread layout of indoor WLAN. However, the signal fluctuations in the complex environments make it difficult to maintain high accuracy localization for the received signal strength (RSS) fingerprinting. Various positioning solutions have emerged to address this challenge, either working in stand-alone mode or in collaborative mode. In the former case, the user only utilizes his own RSS observation to request location service, while the latter usually requires information transfer between users. Considering the spatial correlation of wireless signal distribution, we propose an online joint localization scheme (JointLoc) that does not require direct interaction between users. The fact that the signals observed by users in physical proximity characterize the surroundings is used by JointLoc to identify neighboring users for joint localization. Besides this, JointLoc further integrates a novel subset-based localization scheme, thus the influence of anomalous RSS signals is eliminated before making the final location decision. We have fully evaluated the performance of JointLoc in two RSS datasets collected in real environments. Compared with conventional algorithms and the latest ones, results show that JointLoc is robust against signal fluctuations, and achieves good localization accuracy.